mcp-numpy

An MCP server that exposes NumPy functionality

Install

pip install mcp-numpy

Usage

As an MCP Server

To use with Claude Desktop or other MCP clients, add to your mcp.json:

{
  "mcpServers": {
    "mcp-numpy": {
      "command": "mcp-numpy"
    }
  }
}

Available Tools

The server exposes the following NumPy functionality as MCP tools:

Array Creation

• np_array - Create a NumPy array
• np_zeros - Create zeros array
• np_ones - Create ones array
• np_full - Create array filled with value
• np_arange - Create array with range
• np_linspace - Create evenly spaced array
• np_eye - Create identity matrix
• np_diag - Create diagonal array

Array Manipulation

• np_reshape - Reshape array
• np_transpose - Transpose array
• np_concatenate - Concatenate arrays
• np_split - Split array
• np_tile - Tile array
• np_repeat - Repeat elements
• np_squeeze - Remove single-dimensional entries
• np_flatten - Flatten array

Mathematical Operations

• np_sum, np_mean, np_std, np_var - Summary statistics
• np_min, np_max, np_argmin, np_argmax - Min/max operations
• np_dot, np_matmul, np_cross - Matrix operations
• np_trace, np_cumsum, np_cumprod, np_diff - Array operations

Linear Algebra

• np_inv - Matrix inverse
• np_det - Matrix determinant
• np_eig - Eigenvalues and eigenvectors
• np_svd - Singular value decomposition
• np_solve - Solve linear system
• np_linalg_norm - Matrix/vector norm

Random

• np_rand - Random floats
• np_randn - Random normal
• np_randint - Random integers
• np_random_choice - Random choice
• np_shuffle - Shuffle array

Statistics

• np_percentile, np_quantile - Percentiles/quantiles
• np_histogram - Histogram
• np_correlate, np_corrcoef - Correlation

Element-wise Math

• np_add, np_subtract, np_multiply, np_divide - Arithmetic
• np_power, np_mod - Power and modulo
• np_sqrt, np_abs - Basic math
• np_exp, np_log, np_log10 - Logarithms
• np_sin, np_cos, np_tan - Trigonometry
• np_arcsin, np_arccos, np_arctan - Inverse trig
• np_sinh, np_cosh, np_tanh - Hyperbolic

Array Properties

• np_shape, np_ndim, np_size, np_dtype - Properties
• npastype - Type conversion

Development

git clone https://github.com/daedalus/mcp-numpy.git
cd mcp-numpy
pip install -e ".[test]"

# run tests
pytest

# format
ruff format src/ tests/

# lint
ruff check src/ tests/

# type check
mypy src/

mcp-name: io.github.daedalus/mcp-numpy